Use Vectorize when you need to control embeddings, vector indexing, and retrieval for semantic search, recommendations, or RAG. For a managed retrieval pipeline, see AI Search.

Fetch current documentation before implementing. Start with the Vectorize documentation index to discover pages; load only those relevant to the task. Treat the docs as the source of truth for APIs, configuration, models, limits, and pricing.

Task routing

Task	Read
Create an index and connect a Worker	Configuration and Introduction to Vectorize
Insert, update, query, retrieve, or delete vectors	API routes
Generate embeddings, build RAG, or partition tenant data	Patterns
Diagnose missing matches, metadata, or rejected requests	Gotchas

Decisions to make first

Use a consistent embedding model and preprocessing for stored vectors and queries. Matching dimensions alone does not make different models' embeddings compatible.
Choose dimensions from the embedding output and a distance metric appropriate to that model. Changing either requires a new index; check index configuration and scoring semantics before choosing thresholds.
Plan filterable metadata before ingestion. Adding an index later requires re-upserting existing vectors to index that metadata; see metadata filtering.
A namespace partitions search; your application must authorize access and derive tenant scope from trusted identity. See tenant patterns.
Design for asynchronous mutation visibility rather than assuming a completed write is already searchable. See mutation semantics.